bpo-39648:[WIP]updated math.gcd to accept 'n' arguments.

[ananthan-123]

https://bugs.python.org/issue39648

[ananthan-123]

I made clinic and now it is showing checksum error.
What to do?

[mdickinson]

What error are you getting? make clinic runs fine for me on commit 660448e, and has no effect on mathmodule.c.h, so it looks as though everything is up to date here.

[mdickinson]

Ah, I think I understand: after you merge the upstream master in, you'll get conflicts in mathmodule.c.h. You can either restore mathmodule.c.h to either the local version or the upstream version (check out git checkout --ours and git checkout --theirs), or perhaps more simply, I believe it should always be safe to simply delete mathmodule.c.h entirely and then regenerate it using make clinic.

[ananthan-123]

I think the test failure is due to some other problem!!

[mdickinson]

I think the test failure is due to some other problem!!

Yes. You need to initialise g to be the integer zero (it's currently initialised to the NULL pointer, and then that NULL pointer gets passed into the first call to _PyLong_GCD). After that, there are some reference leaks to sort out.

[sweeneyde]

I think the main bug is that there is a difference between the C value 0 (aka a null pointer) and the Python int 0.

[ananthan-123]

Updated the code.
Do we have to decref 'item'.

[mdickinson]

Do we have to decref 'item'.

No: item is effectively a borrowed reference; no need to decref. Having said that, you don't really need a separate item at all. You could replace:

item = args[i];
in = PyNumber_Index(item);

with simply

in = PyNumber_Index(args[i]);

[ananthan-123]

Thanks for the review.Made suitable changes.

[ananthan-123]

Now the problem is with clinic(don't know if any other problem is there).Don't we have to make clinic.
Do we have to regain previous clinic file.

[mdickinson]

Closing, since #18604 was merged.